On disasters evacuation modeling: From disruptive to slow-response decisions

Decision-making evacuation modeling for natural hazards has always been complex due to the uncertainty of variables affecting it. One of the most relevant existing assumptions for this kind of simulation is the use of sigmoid distributions (e.g., Rayleigh) to represent the evacuation times. We found some disadvantages in using this type of probability distribution for slow-response disasters (i.e., geological hazards applications), where there may be a non-negligible delay time between the beginning of the disaster and the consequent evacuation. For that reason, we developed and fitted a displaced sigmoid distribution to the results of a landslide evacuation simulation, i.e., applied to the 2008 disaster of Chaitén in southern Chile. We considered a Rayleigh distribution with a shape factor and a (proposed) displacement factor for predictions. The results suggest considering a displacement when using the Rayleigh distribution to represent slow-response evacuations. We also provide a detailed discussion about the time-dependent factors that may influence this displacement.